Skip to content

🌱 Top up v6 with v5 (durable hosted hive.yaml, held-gate pills, stale CI chevron re-list and 117 more) - #17570

Merged
clubanderson merged 169 commits into
v6from
topup/v6-from-v5-20261110
Nov 10, 2026
Merged

clubanderson merged 169 commits into
v6from
topup/v6-from-v5-20261110

Conversation

@clubanderson

@clubanderson clubanderson commented Nov 10, 2026 •

Copy link
Copy Markdown
Member

Forward-merge of origin/v5 into v6 (120 non-merge v5 commits, through v5.211.18). Plain --no-ff merge commits — merge with --merge, not squash, so the v5 history is carried.

Supersedes #17506 (stuck: needs-rebase against v6 and red on the v6-only tests its merge resolution did not reconcile). This branch reuses #17506's conflict resolution: it merges #17506's head (6c03b7b, v5 as of 235099b00) onto today's v6, then merges the rest of origin/v5, then fixes the drift.

Conflict resolutions

Merge 1 (6c03b7b → v6), 1 file:

Merge 2 (origin/v5 → result), 18 files:

  • .github/scripts/integration-governance.cjs — TRUSTED_SCRIPTS_TREE set to this head's git rev-parse HEAD:src/scripts (7c365d4a…).
  • CHANGELOG.md — both sides kept: v6.x sections first, then the v5.211.x sections.
  • src/pkg/config/config.go — v6 (v6 split the monolith into per-area files); v5's one change (🐛 fix(dashboard): CI chevron no longer stays lit after every check has concluded #17509 ci_activity_stale_m / StageTrackerCIActivityStale) ported into src/pkg/config/dashboard_config.go.
  • src/pkg/github/automerge/automerge_contributor_lane.go — v5's requestProwPresubmitsForFork block ([scanner] 🐛 prow: post /ok-to-test once when the hive opens or pushes to a non-member fork PR #17474) + v6's contributorHeldRunsReason(reason) condition.
  • src/pkg/github/automerge/automerge_contributor_lane_test.go — v5's /statuses fixture case + v6's recheckHeadSHA / noCIEvidence check-runs case.
  • src/pkg/dashboard/static-src/js/tracker-row-strip.js, static/index.html (stageTrackerChipsHtml) — v5's opts.gate signature and gate-mod filtering + v6's 🧑 needs a person finisher chip (the chip steps aside when the gate pill is rendered).
  • src/pkg/dashboard/static-src/js/tracker-actions.js, static/index.html (action labels) — v5's gate:* verbs + v6's blocker:handoff.
  • src/docs/dashboard-sections/activity-monitor.md — v5's item 5 (gate pill wording) + v6's item 6 (review-swarm-off statements).
  • src/docs/github-app-setup.md — v5's Dependabot-alerts row ([scanner] sec-check loses Dependabot alerts read when run above ISSUES_AND_PRS (TokenTierForRole only refines the contributor rung) #17496) + v6's Code-scanning-alerts row.
  • src/docs/api-reference/{agents-and-controls,beads,configuration,dashboard-misc,inception,knowledge,nous,packs-and-acmm}.md — regenerated with check-api-reference-citations.sh --fix (460 citations rewritten, 0 unresolved).

Test / guard fixes (the v6-only tests #17506 left red)

v6 adopts v5 #17327/#17334: a hold-class label or sentinel alert is a gate riding the head's real stage as a pill, not an 085 park.

  • pkg/stagelabels TestDecisionCommentNamesActionsPerLane — hold and sentinel cases assert the 090 route with Gates and ReadyComment's Held: … wording (no approval offered); routing itself is pinned by the carried TestRouteHumanGates.
  • pkg/github/automerge TestManualMergeNamedWhileHoldLabelBlocks (090 held), TestManualMergeNotClaimedWithoutWorkflowFence/* and TestWorkflowFenceScopeStageNoteAgreesWithLane/unlisted_workflow (071 — the normalized requires_human remediation is the author's, no 085 note) via a manualMergeSweepTo(wantStage, …) helper; other callers unchanged.
  • pkg/policies TestLanePoliciesParkOwnerOnlyHandOffs — redteam-holdgated.md (defaults + src/policies mirror) carries the owner-only park rule block from quality-holdgated.md.
  • changelog-fragment-guard / relative links and anchors resolve — api-reference citations resynced; scale-envelope.md RankActionableIssues range; every v5 fragment was already compiled into v5 releases, so changelog.d/changed-v6-topup-from-v5-20261110.md is added.
  • native-queue-policy / probot DCO — expected red, as on 🌱 top-up: forward-merge v5 into v6 (v6 release line cut, v5.187–v5.189) #17118: probot flags carried v5 commits whose author name differs from the sign-off name; not repairable without rewriting v5 history. Every commit new to this branch is -s signed. Land with --merge --admin as 🌱 top-up: forward-merge v5 into v6 (v6 release line cut, v5.187–v5.189) #17118 was.

Local: go test ./pkg/stagelabels/... ./pkg/github/automerge/... ./pkg/policies/... ./pkg/config/... ./pkg/stageengine/... green; link/citation checks green; check-changelog-fragment.sh v6 OK.

Re-sync with v6 (2026-11-10, after v6.12.1 / #17568)

  • Merged origin/v6 (958e9ca85e, signed). 1 conflict: src/pkg/stageengine/derive_pr_classify.go — kept the v5 stageHeldGateReasons block and re-added 🐛 stageengine: classify repo-auto-merge-disabled from head evidence instead of handing off #17568's new repo-auto-merge-disabled block after it (plus its stageHandoffReasons removal). go test ./pkg/stageengine/... ./pkg/stagelabels/... green.
  • TestStageTrackerRows: the v6 "needs a person" chip now renders beside the v5 human-gate pill; the v6-only hold/needs-human fixtures expect the gate pill instead of per-modifier chips.

DCO / native-queue-policy

probot DCO flags 36 carried v5 commits (author Andy Anderson ≠ sign-off Andrew Anderson, same email). native-queue-policy fails only because of that, and its prow-DCO substitute applies only to sync/v5-to-v6 bot PRs. DCO_WAIVED_COMMITS feeds the post-merge monitor, not probot. Same state as #17118 — land with gh pr merge --merge --admin.

Karibhou and others added 30 commits November 8, 2026 20:23
…17274)

validateWorkspace cleaned workspace_path and compared it to
workspace_roots as a string prefix, then os.Stat followed symlinks. A
link under a root (<root>/x -> /), which a previous run can plant in its
read-write workspace, passed the check and the backend mounted its
target read-write.

The path is now resolved with filepath.EvalSymlinks and the resolved
form is checked against the resolved roots (so a root reached through a
link still works); a path resolving outside is refused 403. The run
mounts and records the resolved directory, so the backend never follows
a link the check did not see.

Fixes #17029

Hive-Run: #17029
Hive-Plan: review-backlog/security
Hive-Spec: crux-runs#workspace-roots

Signed-off-by: Karibhou <Karibhou@users.noreply.github.com>
Co-authored-by: Karibhou <Karibhou@users.noreply.github.com>
Co-authored-by: hivecommons-hive[bot] <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Automated release commit. Compiles changelog.d/ fragments and moves
the CHANGELOG.md Unreleased section into a dated v5.205.0 entry.
See src/docs/releases.md.

Signed-off-by: hive-release-bot <actions@github.com>
Automated release commit. Compiles changelog.d/ fragments and moves
the CHANGELOG.md Unreleased section into a dated v5.206.0 entry.
See src/docs/releases.md.

Signed-off-by: hive-release-bot <actions@github.com>
* 🐛 crux: confine agentstep workspace file access with os.Root

agentstep read and wrote step files on the host with plain os calls after only a lexical '..' check, so a symlink in the workspace (carried by a provisioned repo, or planted by the sandboxed agent for answer files) redirected host-side reads/writes as the serve uid. All prompt_path reads, prompt writes and answer reads now go through an os.Root opened on the workspace, as review_request_sandbox_spool.go and runs.validateWorkspace do.

Closes #17293

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* 🐛 crux: create .crux parent before symlinking steps dir in agentstep test

TestSymlinkedCruxDirCannotRedirectThePromptWrite symlinked .crux/steps without creating .crux first, so os.Symlink failed with ENOENT before the assertion ran.

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

---------

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Automated release commit. Compiles changelog.d/ fragments and moves
the CHANGELOG.md Unreleased section into a dated v5.206.1 entry.
See src/docs/releases.md.

Signed-off-by: hive-release-bot <actions@github.com>
)

* 🐛 hold guard: log deferred base refresh at debug instead of WARN

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* 🐛 hold guard: scope base-refresh deferral to behind so dirty held PRs keep their conflict notice

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

---------

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
…hs beside a conflict (#17304)

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Automated release commit. Compiles changelog.d/ fragments and moves
the CHANGELOG.md Unreleased section into a dated v5.206.2 entry.
See src/docs/releases.md.

Signed-off-by: hive-release-bot <actions@github.com>
…7305)

* 🐛 contribute: convert a PR-covered contributor claim instead of releasing it on lease revoke

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* 📖 api-reference: refresh claims_api.go citations after lease-revoke change

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* 🐛 contribute: hold the claim from the verified task_complete PR, not the scan ledger

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

---------

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Automated release commit. Compiles changelog.d/ fragments and moves
the CHANGELOG.md Unreleased section into a dated v5.206.3 entry.
See src/docs/releases.md.

Signed-off-by: hive-release-bot <actions@github.com>
…uest (#17306)

* ✨ hive-open-issue: --claim files and claims a self-filed issue in one request

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* 🌱 test: decode the bare label array go-github posts in the create-claim fake

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* fix: route the --claim help-wanted strip through the shared blocked-sweep helper

TestDirectWriteRatchet failed on CI: the --claim path added an eighth
wrapper write (RemoveLabel) to issue_request_watcher.go, whose baseline
is 7 and may never be raised. Generalize the blocked sweep's
stripBlockedHelpWanted into stripHelpWanted in issue_blocked.go (where
that removal is already baselined) and call it from claimCreatedIssue, so
the watcher file gains no new direct write and the single help-wanted
removal the claim test asserts still happens.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* 🐛 hive-open-issue: refuse the filer on hub dispatch, bind the marker to the filer, ledger the --claim

Review findings on #17306 (#17299):

- The hub's own scanner dispatch (installClaimAdmissionCheck) still called
  the agent-less IssueDispatchBlockReason, so an issue a lane filed without
  --claim was handed back to that lane — the #17243→#17250 shape. A
  KindAgent request now goes through IssueStartBlockReason with req.Holder
  (the agent name), and a test asserts the filer's dispatch is refused with
  ClaimBlockReasonFiledForHelpPrefix while another agent is admitted.
- withFiledForHelpMarker left a body alone when it already carried any
  marker, so a filer could seed `<!-- hive:filed-for-help other -->` and
  never be bound. Any pre-existing marker is now dropped before the filer's
  own is appended; tested with a seeded foreign marker, and the agent name
  is validated against a named pattern instead of round-tripping through the
  marker regex.
- claimCreatedIssue applied the labels and facts but never raised the agent
  start signal, so the hub's claim ledger was not written for a --claim
  create. It now fires AgentStartSignalClaim like a `claim` request; the
  create-claim test records the hook and asserts one call for #99.
- Docs: the filed-for-help refusal sentence sits beside the other `claim`
  refusals; the filing section notes the dispatch refusal, marker
  replacement and ledger signal.

Hive-Fix-Attempt: 1/6
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* 🐛 hive-open-issue: ledger the --claim via a filed-claim start signal the hub lists

A self-filed issue was never in a kick, so the plain `claim` start signal
the create+claim path raised went through ClaimOnStart with started=false
and the hub's ledger never recorded the claim: dispatch admission Lookup,
the PR-request gate and the claim listing all saw it unclaimed. The create
path now raises AgentStartSignalFiledClaim; recordAgentStart treats that
signal as the issue's own listing (MarkListed, then ClaimOnStart through
the normal admission check). Ledger-backed test in cmd/hive.

Hive-Fix-Attempt: 2/6
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* 🌱 hive-open-issue: document the --claim allowlist exception, share the help-wanted strip, keep the marker off fold comments

The write-surface catalogue said same-path side effects ride the parent
operation's allowlist entry, but the claim performed on --claim is gated
on its own claim allowlist and audited as agent_issue_claimed; the
create_issue row and the side-effects paragraph now say so, and the
hive-open-issue doc names the allowlist requirement.

The --blocked-on-pr comment path removed help wanted directly, so
stripHelpWanted's single-call-site claim was false; it now goes through
the helper.

The filed-for-help marker was appended before consolidation and so landed
in file-set, component and subject fold comments on someone else's issue,
where nothing reads it. The fold renderers strip it.

Hive-Fix-Attempt: 3/6
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* 🐛 hive-open-issue: exempt --parent children from the filed-for-help marker, collapse blank lines only on strip

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

---------

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…er author (#17303)

* 🐛 pr-request: refuse Fixes/Closes targets claimed or covered by another author

A PR request could close an issue a contributor had claimed and covered
with their own open PR (#17243: #17250 superseded #17248). The watcher
now checks each closing reference before booking a start: a live claim
by another principal, or an open covering PR by an author outside the
hive, rejects the request (gate claim) and posts one marker-deduplicated
notice on the issue. A failed lookup fails closed.

Closes #17298

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* 🐛 pr-request: own-claim exemption checks claim kind and hive; scope claim notice to served repos

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* 🌱 test: cover the closed-covering-PR release and the forge-assignee claim read

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* 🐛 issue-request: run the claim-target gate on the start path before notifyAgentStart

Review of #17303 (intent-alignment): #17298's acceptance criteria require
the kick/start path to run the claim/coverage check before a start is
booked, but only the PR-open leg was gated. A claim or a start-announcing
relay comment on an issue another principal has claimed, or covered with
an open PR, now goes through the same prRequestTargetBlock before any
write and before notifyAgentStart: refused with a policy reason and the
deduplicated notice, retried on a lookup error (fail closed). The retry
path is extracted into failIssueRequest so both callers share it.

Hive-Fix-Attempt: 1/6
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* 🐛 pr-request: name the refused action in the claim-target notice

The notice posted on a refused target issue hard-coded "asked to open a
pull request ... No PR was opened", so a refused `hive-open-issue claim`
or start comment described an event that never happened. Thread the
action (open-pr / claim / start) into the notice body, assert the
start-path wording in TestIssueRequestStart_RefusesContributorClaimedTarget,
and document the new refusal reason in the hive-open-issue `claim` section.

Hive-Fix-Attempt: 2/6
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* 🌱 test: serve the comments list in the in-progress relay fixture

With issue claims on, the start-path claim-target gate reads the forge
marker from the issue's comments before applying the claim; the fixture
404'd that read, so the gate failed closed and
TestIssueRequestWatcher_ClaimLeavesCommentToLedgerWhenClaimsOn lost its
ownership label.

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* 🐛 pr-request: claim-target gate rejects a missing target and skips pull requests

With issue claims on, a closing reference to a nonexistent issue 404ed
in prRequestTargetClaim and came back as a plain error, so the request
was failed and retried until the horizon quarantined it; before #17298
the same request was permanently rejected with the cross-repo hint. The
404 is now mapped to that policy error in validatePRRequestTargets.

The start-comment gate also fetched PRs via Issues.Get and inferred a
claim from assignees, refusing a hive start comment on a contributor-
assigned PR. A target that is a pull request now passes the gate.

Docs: name the dispatch-block gate explicitly where the inserted
claim-target paragraph left 'the same gate' ambiguous.

Hive-Fix-Attempt: 3/6
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

---------

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Automated release commit. Compiles changelog.d/ fragments and moves
the CHANGELOG.md Unreleased section into a dated v5.207.0 entry.
See src/docs/releases.md.

Signed-off-by: hive-release-bot <actions@github.com>
Automated release commit. Compiles changelog.d/ fragments and moves
the CHANGELOG.md Unreleased section into a dated v5.207.1 entry.
See src/docs/releases.md.

Signed-off-by: hive-release-bot <actions@github.com>
…17321)

The scheduler's injection-cap WARN is the operator-side notice that a repository's AGENTS.md outgrew the cap (#12129): repo content, the cap working as designed, not a hive defect. Every kick re-logs it, so one oversized repo crossed the repeat threshold and was filed as an unclassified error (#17318). Exempt the exact message from the feed; 'agentsmd: parse failed, skipping injection' stays reportable.

The warning also printed cap_bytes=0, the config 'use default' sentinel. Log the effective cap instead.

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
…ash (#17323)

* 🐛 watchdog: stop paging PaneShowsLogin on a relaunched CLI's login flash

A Copilot CLI relaunched mid-life (kick restart, memory-drain recycle,
watchdog restart) paints login chrome for ~0.5s while it re-validates its
token. Classify dated boot grace from the agent's original StartedAt, so a
22h-old agent got no grace and a single sighting flipped
Authenticated=False/PaneShowsLogin and paged the operator.

- Observation.LastLaunchAt records the most recent CLI spawn
  (markCLILaunchedLocked); boot grace uses the later of StartedAt and it.
- ApplyLoginStreak holds auth-required back until login chrome has been
  seen on Settings.LoginStreak consecutive polls (liveness.login_streak,
  default 3), mirroring the pane poller's loginStreak guard. Below the
  threshold the pane is unknown with reason 'login chrome seen 1/3 polls'.

Fixes #17320

Signed-off-by: Andrew Anderson <andy@clubanderson.com>

* test: add Liveness.LoginStreak to the agent-runtime config golden

Signed-off-by: Andrew Anderson <andy@clubanderson.com>

* dashboard: reference governor.watchdog.liveness.login_streak in the Watchdog settings help

Signed-off-by: Andrew Anderson <andy@clubanderson.com>

---------

Signed-off-by: Andrew Anderson <andy@clubanderson.com>
…7326)

The popover opened from the Merged pill was capped at 34rem, so every
PR title ellipsised at ~40 characters. Set it to 52rem, capped to the
viewport and to the sticky bar that clips it (it is right-anchored
under the pill and grows leftward). The title column is the grid's only
flexible track, so the width goes to titles; the by/lane/time meta
column keeps its natural nowrap width. The merge-queue popover keeps its
old 22–34rem sizing and the ≤720px phone sheet is a separate dialog and
is unchanged.

Signed-off-by: Andrew Anderson <andy@clubanderson.com>
#17325)

A PR closing two issues is drawn once, under the lower issue; the other
issue chips to it. Its row fell back to the issue's own 030/040 stage for
the lane while the merge-gate word still came from the PR, so #17302 read
"Soaking · 47s left" in the PR in progress lane while #17292 soaked in
the Merge queue. stageTrackerLinkedRiding picks the furthest covering PR
drawn elsewhere and the row, column map and sort order ride its stage;
the PR column stays empty and the "under #M" chip still names the PR.

Fixes #17324

Signed-off-by: Andrew Anderson <andy@clubanderson.com>
Automated release commit. Compiles changelog.d/ fragments and moves
the CHANGELOG.md Unreleased section into a dated v5.207.2 entry.
See src/docs/releases.md.

Signed-off-by: hive-release-bot <actions@github.com>
#17331)

* 🐛 contributor nudge: per-PR CI-failure cooldown and failing-set dedupe

CI-failure nudges are keyed on the failing-check set rather than the lane's
reason prefix. The same set on a new head, or a check joining the set on the
same head, edits the PR's one nudge comment (marker gains checks=<digest>
attempt=<n>; no relay re-dispatch for the same head). A changed set on a new
head within governor.labels.stage.contributor_nudge.ci_failure_cooldown
(default 6h) is folded into the comment instead of counting as a fresh
nudge. The ledger persists in the stage-label store across restarts.

Fixes #17328

Signed-off-by: Andrew Anderson <andy@clubanderson.com>

* test(config): golden gains contributor_nudge.ci_failure_cooldown

Signed-off-by: Andrew Anderson <andy@clubanderson.com>

---------

Signed-off-by: Andrew Anderson <andy@clubanderson.com>
…stage instead of forcing needs-decision (#17334)

Sentinel alerts and hold-class labels are gates, not stages (owner decision
2026-11-08). The stage engine now routes a gated PR by its head's evidence
like any other human hand-off — CI / 070 while checks run or fail, 071 on a
reviewer question, 090-ready-for-human-merge when the head is otherwise
clean — and keeps the gate reason on the route so the entry comment reads
"Held: a sentinel alert is open on this PR. The hive never merges this head
while the gate stands." and the dashboard popover still explains the stop.

The gate remains a hard merge stop: the automerge lanes and the dashboard's
delegated human-merge shortcuts refuse on the label itself, never on the
stage number (tests pin a 090 PR with a sentinel alert or hold refused by
the contributor, self-authored and trusted-author prefilters and by the
dashboard's human-merge decision/outcome paths).

Dashboard: a held 090 row's Merge queue cell reads "Held · sentinel ·
confirm intent" / "Held · hold" (amber ham-held, tooltip with the gate's
reason, the route reason and the rule), never "Queued for merge"; the
sentinel question's Approve / Let hive decide / direction / close actions
stay available from that cell. needs-human keeps parking at 085.

Fixes #17327

Signed-off-by: Andrew Anderson <andy@clubanderson.com>
Co-authored-by: hivecommons-hive[bot] <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Automated release commit. Compiles changelog.d/ fragments and moves
the CHANGELOG.md Unreleased section into a dated v5.207.3 entry.
See src/docs/releases.md.

Signed-off-by: hive-release-bot <actions@github.com>
@kubestellar-prow kubestellar-prow Bot added the dco-signoff: yes Indicates the PR's author has signed the DCO. label Nov 10, 2026
@clubanderson

Copy link
Copy Markdown
Member Author

/lgtm
/approve

@kubestellar-prow

Copy link
Copy Markdown
Contributor

@clubanderson: you cannot LGTM your own PR.

Details

In response to this:

/lgtm
/approve

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository.

@kubestellar-prow kubestellar-prow Bot added approved Indicates a PR has been approved by an approver from all required OWNERS files. size/XXL Denotes a PR that changes 1000+ lines, ignoring generated files. labels Nov 10, 2026
@github-actions

github-actions Bot commented Nov 10, 2026 •

Copy link
Copy Markdown
Contributor

Open PR overlap check

Base branch: v6

Conflicts are checked with every PR merged onto the current tip of v6, so a PR that only needs a rebase is not reported as conflicting with its siblings.

PR Title Relation Files
#17576 🐛 watchdog: don't raise CredentialMissing for login chrome inside boot grace same files src/crux/watchdog/reconciler.go
#17572 ✨ celtrigger: ci.failed event for default-branch CI failures same files src/cmd/hive/github_client_config.go
src/docs/api-reference/configuration.md
src/pkg/dashboard/static/index.html
#17329 🐛 [v6] knowledge priming: search on what the issue is about, not hive routing same files src/docs/agent-knowledge-and-memory.md
src/pkg/scheduler/kick_priming.go

@github-actions github-actions Bot added conflicts-with-open-pr Open PR has a merge conflict with another open PR and removed conflicts-with-open-pr Open PR has a merge conflict with another open PR labels Nov 10, 2026
@hivecommons-hive

hivecommons-hive Bot commented Nov 10, 2026 •

Copy link
Copy Markdown
Contributor

Hive status — updated in place; edits do not notify. Full history is in the Activity Monitor.

@clubanderson this PR is held at stage/071-needs-author and the hive will not merge it yet.

Failing checks:

  • native-queue-policy

Next: the hive does not push to your branch. Push a new head (or reply here); CI and the review swarm re-run automatically and the PR returns to the normal path.

Review findings from reviewer (folded here because this head already carries its hive review):

Intent alignment (read at head 958e9ca8): the branch is the 120 carried v5 commits plus ce0fdb7/17f6d35/958e9ca merges and three reconciliation commits (c13107a, 6984a3a, 3d6b0d6). Every reconciliation claim in the body checks out — TRUSTED_SCRIPTS_TREE (.github/scripts/integration-governance.cjs:12) equals the head's src/scripts tree 7c365d4a…; redteam-holdgated.md is v5-only and v6's TestLanePoliciesParkOwnerOnlyHandOffs (src/pkg/policies/needs_decision_filing_test.go:52-77) requires the park block; derive_pr_classify.go vs v6 base differs only by the stageHeldGateReasons block (:310-321, :489-509). No scope beyond the top-up. Two advisories, neither blocking:

  1. Body is self-contradictory on the person chip. "Conflict resolutions" says the 🧑 needs a person chip "steps aside when the gate pill is rendered", but 3d6b0d6 reversed that (src/pkg/dashboard/static-src/js/tracker-row-strip.js:984 is now const isPerson = item.finisher === 'person'; with no !gated), and the later "Re-sync" section says the chip renders beside the pill. Suggest deleting the parenthetical in the first section so the body matches the diff.

  2. Open question, out of scope for the merge: src/pkg/stageengine/derive_pr_classify.go:546 (v6 🐛 stageengine: classify repo-auto-merge-disabled from head evidence instead of handing off #17568's repo-auto-merge-disabled branch) still parks a hold label / sentinel alert at 085, while the gate semantics this PR imports (:319, :489) ride the same hold on the real stage as a pill. If the lane can return repo-auto-merge-disabled for a held head, v6 now has two answers for "hold on a green, approved PR". Not introduced by this merge — worth a follow-up issue rather than a change here.

Review findings from reviewer-2 (folded here because this head already carries its hive review):

Two things block this as-is: TestStageTrackerRows is red for a reason that is in this branch, and the branch now conflicts with v6 in one file. Neither is the DCO/probot noise the body already explains.

1. test (dashboard 1/4) and dashboard shuffle 1/3 — deterministic, 3/3 runs

src/pkg/dashboard/stage_tracker_ui_rows_test.go:903-909 are v6's 🧑 needs a person chip assertions. They expect, for a finisher:'person' row at stage 85 with modifiers:['needs-human','needs-decision'], the chip with the folded pill names in its hover, and (lines 907/909) that hold keeps its own pill and that a finisher:'hive' row still shows ham-mod-needs-human.

The merged renderer no longer does any of that on a gated row — by design, per the body ("the chip steps aside when the gate pill is rendered"):

So the three v6 assertions and the v5 gate-pill design are mutually exclusive; the test was not reconciled when the two were merged. The CI log confirms it's the first of these that trips (the rendered row carries ham-human-gate … Needs human · decision and no ham-finisher-person). The body's "Local: go test ./pkg/stagelabels/... ./pkg/github/automerge/... ./pkg/policies/... ./pkg/config/... ./pkg/stageengine/... green" is accurate but omits ./pkg/dashboard/..., which is where the red is.

Fix is in the test, not the renderer: the #128/#129/#130 rows need to assert the v5 outcome (ham-human-gate present, no chip, no ham-mod-hold/ham-mod-needs-human pills) — or move them to a non-gated stage if the intent is to keep testing the chip's fold. The v6 #120 row at stage 50 (line 898) still passes because stageTrackerHumanGateReasons doesn't flag it.

2. Conflicts with v6 since #17568 landed

git merge v6 into this head conflicts in src/pkg/stageengine/derive_pr_classify.go (hunks at ~489, ~502, ~515): v6 #17568 added the reason == "repo-auto-merge-disabled" branch and kept the SentinelAlert || Holds → NeedsDecision park in the owned-rank path, while this branch replaces that block with stageHeldGateReasons[reason] and the 090-pill route (#17327/#17334). Those are the same hold-vs-park semantics the body describes fixing in pkg/stagelabels and automerge, so the resolution should take the v5 gate design and re-add the #17568 branch alongside it — not a mechanical pick of either side.

Not re-litigated

native-queue-policy / probot DCO red on carried v5 commits is as the body states and as #17118 landed; the --merge --admin landing is clubanderson's call. I did not read the other ~340 files in this merge beyond the conflict-resolution list; nothing above is a claim about them.

hive/awaiting-ci: mergeable=CONFLICTING against v6 at 6984a3a (plus red test/coverage/native-queue-policy and DCO action-required). Human-authored branch — hive will not push here; deferring until the author rebases onto current v6.


🐝 Hive Agent: scanner | Instance: hosted-available-oke-11-placeholder-r05x | SHA: unknown

— hive: agent=scanner backend=copilot model=claude-fable-5.1 copilot=1.0.88

Review findings from adjudicator (folded here because this head already carries its hive review):

Correctness review at 958e9ca8 (merge-base→head diff read via local clone; hand-resolved files also diffed against origin/v5 @ ae32a64e4). The conflict resolutions and the drift fixes match the PR body; CI is green except the DCO/native-queue-policy failures the body already explains.

Advisory (pre-existing, not blocking): src/pkg/stageengine/derive_pr_classify.go:546 — the v6-only repo-auto-merge-disabled branch (#17568) still returns NeedsDecision on o.Detail.SentinelAlert || len(stagelabels.Holds(o.Labels)) > 0, while this merge adopts v5 #17327's gate model everywhere else (stageHeldGateReasons at line 319, stagelabels.withGates). On a repo with auto-merge disabled a held PR will show 085 with no Held · … pill; elsewhere it shows its real stage plus the pill. Worth a follow-up that drops that early return and lets RouteHuman carry the gate, as the stageHeldGateReasons branch above it does.

Verified: TRUSTED_SCRIPTS_TREE equals git rev-parse HEAD:src/scripts (7c365d4a…); StageTrackerCIActivityStale in dashboard_config.go:540 is semantically identical to v5's config.go:5372 and consumed at stage_tracker_build.go:1144; redteam-holdgated.md defaults and src/policies mirror are byte-identical; tracker-row-strip.js differs from v5 by only the v6 person-chip lines.

@hivecommons-hive

Copy link
Copy Markdown
Contributor
6984a3a

ℹ️ Sentinel notice — informational; author is trusted (owner), merge is not blocked

Hive flagged this PR (author @clubanderson, head 6984a3a51bba) because it matches behaviors that can override security controls, escalate privileges or damage the codebase. This is a heuristic, not an accusation — a maintainer should confirm the change is intended before it merges.

  • sensitive_path — changes 20 sensitive path(s) (Touches a sensitive path (OWNERS, workflows, policies, hive config, security docs, …))
    • .github/dependabot.yml
    • .github/release-lines.yml
    • .github/workflows/ci-runner-image.yml
    • .github/workflows/docker.yml
    • .github/workflows/integration-candidate.yml
    • .github/workflows/integration-technical-v2-ci.yml
    • .github/workflows/mutation-advisory.yml
    • .github/workflows/pr-verifier.yml
    • .github/workflows/promote-stable.yml
    • .github/workflows/stream-image.yml
    • .github/workflows/tagged-release.yml
    • .github/workflows/v2-ci.yml
    • src/Dockerfile
    • src/Dockerfile.contributor
    • src/deploy/entrypoint.sh
    • src/deploy/k8s/error-pages.yaml
    • src/pkg/policies/content_boundary_test.go
    • src/pkg/policies/defaults/redteam-holdgated.md
    • src/policies/README.md
    • src/policies/redteam-holdgated.md
  • permission_escalation — widens workflow permissions or uses pull_request_target (Widens GitHub Actions permissions or switches to pull_request_target)
    • .github/workflows/mutation-advisory.yml
  • ci_gate_weakening — CI configuration is deleted, softened or has gate checks removed (Disables or softens CI gates (continue-on-error, || true, removed checks))
    • .github/workflows/mutation-advisory.yml
  • security_policy_edit — edits security policy, scanner config, rulesets or hooks (Edits SECURITY.md, dependabot/CodeQL config, rulesets or git hooks)
    • .github/dependabot.yml

Hive did not add the sentinel-alert label because this author is trusted. The finding remains recorded for audit and the dashboard Security tab; set sentinel.trusted_authors_block: true to require the blocking label for trusted authors too.

@hivecommons-hive hivecommons-hive Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Correctness review (head 6984a3a) — one in-scope regression the reconciliation commit missed.

TestStageTrackerRows is red on test (dashboard 1/4) / test (dashboard shuffle 1/3) — not in the "expected red" list

The merge-2 resolution of stageTrackerChipsHtml (src/pkg/dashboard/static-src/js/tracker-row-strip.js:986, mirrored in static/index.html) deliberately makes the 🧑 needs a person chip step aside when the human-gate pill renders (isPerson = item.finisher === 'person' && !gated) and filters STAGE_TRACKER_GATE_MODS (hold, needs-human, needs-decision, line 788) out of the pill list (lines 994–996). That matches the PR body. But the v6-only block of the node fixture in src/pkg/dashboard/stage_tracker_ui_rows_test.go:902-910 still asserts the pre-gate behaviour:

  • line 905 — expects the chip with also flagged: needs human, needs human decision; the row is a PR at stage 85, so stageTrackerHumanGateReasons (tracker-human-gate.js:84-86, decisionStage → flagged) emits a decision gate, gated is true and the chip is never rendered. CI output confirms: the row carries data-ham-gate-reasons="decision" and no ham-finisher-person.
  • line 908 — expects ham-mod-hold; it is a gate mod and is filtered when gated.
  • line 910 — expects ham-mod-needs-human with finisher: 'hive'; the gate is reason-driven, not finisher-driven, so this pill is filtered too.

The body's local run (pkg/stagelabels pkg/github/automerge pkg/policies pkg/config pkg/stageengine) skipped pkg/dashboard, which is why this was not seen; CI's test, coverage and tide failures follow from it. Suggested reconciliation of the three cases to the behaviour the resolution adopts:

--- a/src/pkg/dashboard/stage_tracker_ui_rows_test.go
+++ b/src/pkg/dashboard/stage_tracker_ui_rows_test.go
@@ -901,11 +901,13 @@
-// The needs-human and needs-decision pills only repeat the chip, so they fold
-// into its hover; any other modifier still gets its pill.
+// At 085 the human-gate pill carries the ask (hivecommons/hive#17327), so the
+// person chip and the gate-reason pills (needs-human, needs-decision, hold)
+// step aside; any other modifier still gets its pill.
 row = stageTrackerRowHtml({ repo: 'acme/console', number: 128, kind: 'pr', stage: 85, finisher: 'person', finisher_verb: 'DECIDE', finisher_reason: 'Waiting on a decision', modifiers: ['needs-human', 'needs-decision'] }, stages, ctx);
-assert.ok(row.includes('title="Hive cannot finish this on its own: DECIDE — Waiting on a decision · also flagged: needs human, needs human decision">🧑 needs a person</span>'), 'folded pills are named on the chip: ' + row);
-assert.ok(!row.includes('ham-mod-needs-human') && !row.includes('ham-mod-needs-decision'), 'no repeated pill beside the chip: ' + row);
+assert.ok(row.includes('data-ham-gate-reasons="decision"') && !row.includes('ham-finisher-person'), 'the gate pill replaces the person chip: ' + row);
+assert.ok(!row.includes('ham-mod-needs-human') && !row.includes('ham-mod-needs-decision'), 'no repeated pill beside the gate: ' + row);
 row = stageTrackerRowHtml({ repo: 'acme/console', number: 129, kind: 'pr', stage: 85, finisher: 'person', modifiers: ['needs-human', 'hold'] }, stages, ctx);
-assert.ok(row.includes('ham-mod-hold') && !row.includes('ham-mod-needs-human') && !row.includes(' +1'), 'only the repeated pill folds; hold keeps its own, with no +1 for the folded one: ' + row);
+assert.ok(/data-ham-gate-reasons="[^"]*hold/.test(row) && !row.includes('ham-mod-hold') && !row.includes('ham-mod-needs-human') && !row.includes(' +1'), 'hold rides the gate pill, not its own pill: ' + row);
 row = stageTrackerRowHtml({ repo: 'acme/console', number: 130, kind: 'pr', stage: 85, finisher: 'hive', modifiers: ['needs-human'] }, stages, ctx);
-assert.ok(row.includes('ham-mod-needs-human') && !row.includes('ham-finisher-person'), 'without the chip the needs-human pill stays: ' + row);
+assert.ok(row.includes('data-ham-gate-reasons="decision"') && !row.includes('ham-mod-needs-human') && !row.includes('ham-finisher-person'), 'the gate is reason-driven, not finisher-driven: ' + row);

If the intent is instead to keep the chip+folded-pill behaviour for 085 rows, the resolution at tracker-row-strip.js:986 (and index.html) is what needs to change, not the test — but the PR body describes the gated behaviour, so I read the test as the drift.

Everything else I checked in the manual resolutions reads faithfully: automerge_contributor_lane.go:1645-1652 (v5 requestProwPresubmitsForFork + v6 contributorHeldRunsReason), the #17509 port into dashboard_config.go (field, default and StageTrackerCIActivityStale identical to v5's config.go hunk), TRUSTED_SCRIPTS_TREE matches git rev-parse HEAD:src/scripts, and static/index.html mirrors the static-src resolution.

— hive: agent=adjudicator backend=copilot model=claude-fable-5.1 copilot=1.0.88

@hivecommons-hive hivecommons-hive Bot added the needs-rebase Indicates a PR cannot be merged because it has merge conflicts with HEAD. label Nov 10, 2026
The v6 'who finishes it' chip and the v5 human-gate pill both render on a
gated row; the hold/needs-human fixtures now expect the gate pill instead
of per-modifier chips.

Signed-off-by: Andrew Anderson <andy@clubanderson.com>

@hivecommons-hive hivecommons-hive Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Correctness — delta re-review at 3d6b0d6 (vs 6984a3a)

Re-check of prior notes:

  • Fixed — stage_tracker_ui_rows_test.go:905-910 finisher-chip vs gate-pill assertions: the #129/#130 assertions now expect ham-human-gate + Needs human · decision[, hold], and tracker-row-strip.js:984 drops && !gated so #128's folded-chip assertion (:904-905) can pass. Read-only verified; I could not run ./pkg/dashboard (blocked in this pod), and CI's build/merge jobs were skipped on this head, so the package has still not been exercised by CI.
  • Still present (blocking) — src/pkg/stageengine/derive_pr_classify.go conflicts with v6. gh pr view reports mergeable: CONFLICTING, and git merge origin/v6 on this head stops at CONFLICT (content): Merge conflict in src/pkg/stageengine/derive_pr_classify.go. The head's workflow run skipped every build/merge job, so nothing downstream has been validated. Please rebase onto current v6 and resolve.

New, advisory (low): with isPerson = item.finisher === 'person' (tracker-row-strip.js:984, mirrored at index.html:19894), a gated person row now renders both the 🧑 needs a person chip (:988-989) and the human-gate pill (tracker-row-render.js:209 / tracker-row-layout.js:279 pass { gate: !!gate } and append the pill). The comment still above it at :980-982 says the pills fold into the chip precisely to avoid saying "needs a person" twice; the deleted comment said the chip steps aside when the gate pill carries the ask. Not a correctness bug, but the comment and behavior now disagree — either update the comment to say the chip is kept beside the gate pill intentionally, or restore the v5 behavior and change #128's assertion to expect the gate pill instead.

— hive: agent=adjudicator backend=copilot model=claude-fable-5.1 copilot=1.0.88

Signed-off-by: Andrew Anderson <andy@clubanderson.com>

# Conflicts:
#	src/pkg/stageengine/derive_pr_classify.go
@kubestellar-prow kubestellar-prow Bot removed the needs-rebase Indicates a PR cannot be merged because it has merge conflicts with HEAD. label Nov 10, 2026
@clubanderson

Copy link
Copy Markdown
Member Author

/lgtm
/approve

@kubestellar-prow

Copy link
Copy Markdown
Contributor

@clubanderson: you cannot LGTM your own PR.

Details

In response to this:

/lgtm
/approve

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository.

@kubestellar-prow

Copy link
Copy Markdown
Contributor

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: clubanderson

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Details Needs approval from an approver in each of these files:

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

@clubanderson
clubanderson merged commit e7b6bd4 into v6 Nov 10, 2026
100 of 108 checks passed
@kubestellar-prow
kubestellar-prow Bot deleted the topup/v6-from-v5-20261110 branch November 10, 2026 18:53
@github-actions

Copy link
Copy Markdown
Contributor

Thank you for your contribution! Your PR has been merged.

We'd love to hear how your experience was: share feedback

@hivecommons-hive hivecommons-hive Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Security perspective — no blockers (reviewed head 958e9ca8 vs merge-base 3719425d; PR already merged as e7b6bd4).

Verified: .github/scripts/integration-governance.cjs:12 TRUSTED_SCRIPTS_TREE matches git rev-parse 958e9ca8:src/scripts; every docker/login-action@8aa5831f pin resolves to tag v4.8.0; src/crux/runs/service.go:466-485 closes a real symlink escape in validateWorkspace; /v1/health (serve.go:176) stays behind authenticate (serve.go:190).

One low advisory (follow-up, not blocking):

.github/workflows/mutation-advisory.yml:27-29 — the "run" step is not actually token-free. The comment at lines 37-39 says the PR-controlled mutation run "gets no token", but actions/checkout defaults to persist-credentials: true, so the job token (contents: read, pull-requests: write, line 23-25) sits in .git/config while gremlins executes PR test code. Impact is low — the trigger is pull_request, so fork PRs get a read-only token and same-repo authors already hold write — but the stated isolation is incomplete and mutation-report.sh:47 only checks the env var.

      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          fetch-depth: 0
          persist-credentials: false

mutation-report.sh post uses gh with GH_TOKEN from the env, so it does not need the persisted git credential.

— hive: agent=ci-maintainer backend=copilot model=claude-fable-5.1 copilot=1.0.88

@hivecommons-hive

hivecommons-hive Bot commented Nov 10, 2026 •

Copy link
Copy Markdown
Contributor

Review follow-ups (below review.backlog.min_severity: P2)

Non-blocking findings below the backlog filing line are listed here instead of as issues, one comment per PR, updated in place. Tick them off as they are fixed; set review.backlog.min_severity: P3 to file every finding.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

approved Indicates a PR has been approved by an approver from all required OWNERS files. dco-signoff: yes Indicates the PR's author has signed the DCO. size/XXL Denotes a PR that changes 1000+ lines, ignoring generated files.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants